evocation/README.txt

       .   ~     .
       ~   ~~   .
      ~~  ~~~
    ~ ~~~ ~~~~  ~
   ~~~~~~~ ~~~~~
  ~~~~~~~~~~~~~~ .
  ~~ Evocation! ~~
  ~~~~~~~~~~~~~~~             or, how to call the blue-green flame
  ~~~~~~~~~~~~~~
   ~~~~~~~~~~~~~~~

  Evocation is a dialect of Forth, grown to Irenes' tastes. It is meant to
someday be a platform for experimenting with parse theory, type theory,
databases, and other things Forth is not traditionally known for, as well as
with language design, which it is. It is a self-hosting compiler, meaning the
only thing you need to build it is a copy of itself. It is written entirely
without the aid of differentiable neural networks in any capacity, and always
will be.

  Evocation has practical utility, but the primary reason for making it is
spiritual and aesthetic: Irenes believe it is a thing that should exist, and
chose to create it in the most enjoyable way they could find.

  At present, Evocation targets only one architecture, amd64. It is rare among
compiled Forths in that it targets a 64-bit architecture.

  In addition to being self-hosting, Evocation is also self-bootstrapping,
meaning that the compiler can run in a special "show your work" mode which,
instead of a regular binary, outputs a commented hex dump which is heavily
instrumented with explanations of the provenance of every single byte of
output and how to audit that it is correct. This hex dump is suitable for
checking into source control, and can then be converted to a binary executable
through a process that is itself straightforward to audit.

  This hex-dump strategy rests on the insight, from the mescc and guix
developers, that the difference between source code and binary is comments.
However, to Irenes' knowledge, Evocation is the first compiler to be able to
generate this hex dump of itself from an implementation in a high-level
language, rather than an assembly language. Thus the coinage of the term
"self-bootstrapping" to describe it.

  Those not familiar with Ken Thompson's classic paper on this topic[1] will
find it a short, lighthearted read which provides the motivation for this type
of bootstrapping.

  The details of how the bootstrapping is used are described below under
"Hexing Evocation for Distribution".

  This documentation is a perpetual work in progress, but it's getting closer
to being a viable introduction. Please do weigh in on things you wish it
talked about but doesn't.

  If you've obtained this code repository without a link, its authoritative
copy is at https://code.irenes.space/evocation - but there's no need to hit
the network; everything you need should already be in this directory.


 ~~~~~~~~~~~~
~~ Building ~~
 ~~~~~~~~~~~~

  Since we have chosen not to distribute Evocation in binary form, you don't
have a copy of it yet and cannot take advantage of its self-hosting properties
for your first-ever version. Happily, because it is also self-bootstrapping,
you don't need to. Simply run:

  $ ./hex < evoke.hex > evoke
  $ chmod 755 evoke

  These are Unix shell commands, which need to be run in a command terminal
of some sort. If you're running Linux or another Unix variant, you probably
already have a program that gives you a terminal; there are many to choose
from. On other OSes, you may need to go find one. Using the terminal may be
scary at first, but once it becomes familiar it is a very deep creative tool
that changes how you relate to computers. We promise to be gentle about not
demanding too much proficiency, too quickly.

  If you need to learn about the terminal, you may enjoy the Fujoshi Guide to
Web Development[2] or the Wizard Zines[3].

  This tiny program "hex" has been checked into git as a binary, and is the
root of trust for everything Evocation does. It is the only binary component.
If you wish, you may inspect its contents by any means you wish and compare
them to the bootstrapped hex dump of "hex" itself, in the file "hex.hex",
which describes the purpose and provenance of every byte. It's short.

  It's a significantly larger undertaking, but if you have sufficient reason
to, you can audit the contents of "evoke.hex" in the same way.

  Now keep your "evoke" binary somewhere safe, and use it to build new
versions as you modify Evocation.


 ~~~~~~~~~~~~~
~~ Exploring ~~
 ~~~~~~~~~~~~~

  You can now try out Evocation. Type to it interactively:

  $ ./evoke
  ." Hi, Irenes!"
  6 7 * . newline
  bye

  See what it prints!

TODO give examples of RPN for arithmetic

  Some helpful words to try to get started are list-dictionary and describe.
TODO show how to use them

  The syntax for a string literal is s" ...". There's something very subtle
happening: it's the lowercase letter "s", a double quote, and a space. Then
you type the actual contents of the string, then at the end you type another
double quote. The words "Hi, Irenes!", in the example above, are part of a
closely related syntax that uses a period instead of a letter s, and prints
the string out to your terminal instead of returning it.

  Whether you're using s" or .", that space after the quote is mandatory,
which may seem very strange if you're more familiar with pretty much any
language that isn't a Forth, but it's a common Forth idiom. Requiring the
space allows the string literal syntax to be tokenized just like any other
space-delimited word. Unlike most languages, there's no special concept of an
operator or punctuation character that can "interrupt" another word or run up
against the start of one. Everything is separated by spaces.

  Of course, in modern Forth dialects it's also very common to add a special
lexer feature for that sort of thing. Evocation doesn't do that, because
eventually fancy syntax and grammar will be implemented at a higher layer,
using the experimental parsing formalism that doesn't exist yet. If you want
to see where this feature would be if it were implemented in the simpler way,
you can read the definition of the word named "word", in interpret.e.

  If you went to look at that and you're wondering: Yeah, Evocation's lexer
really is that short and simple. Part of why it's able to be that easy is that
words such as s" that introduce special syntaxes do their own lexing for
whatever comes after them.

TODO show how to define words

  Evocation has high-level flow-control words: if, unless, if-else, forever,
and while. High-level flow control is a common thing for modern Forth dialects
to add, but every dialect does it a bit differently. Evocation's flow-control
words are postfix operations and work with curly braces, like this:

  $ ./evoke
  : count 10 0 { 2dup < } { space dup . 1+ } while 2drop newline ;
  count

  What will it print? :)

  Evocation's high-level flow control works only in compiled code; this
example defines and compiles a new word called "count", in order to show it
off. If you try to use the "{ ... } { ... } while" syntax outside of a word
definition, it won't do what you expect.

  This is because, unlike some modern Forths, Evocation doesn't have a
general-purpose memory management facility; it uses something called the log,
which makes it easy to allocate things but hard to deallocate them. In order
to loop through a code block, it has to be allocated somewhere. So, the design
takes care not to encourage programming habits that would burn through memory
space.


 ~~~~~~~~~~~~~~~~~~~~~
~~ Advanced features ~~
 ~~~~~~~~~~~~~~~~~~~~~

TODO where should this go? should there be an interactive tutorial?

: baz ." baz baz baz!" newline 1 nexit ;
: bar ." bar start" newline baz baz baz ." bar end" newline ;
: foo ." foo start" newline bar ." foo end" newline ;
foo
foo start
bar start
baz baz baz!
baz baz baz!
baz baz baz!
bar end
foo end
: baz ." baz baz baz!" newline 2 nexit ;
: bar ." bar start" newline baz baz baz ." bar end" newline ;
: foo ." foo start" newline bar ." foo end" newline ;
foo
foo start
bar start
baz baz baz!


 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~ Reading Evocation's source code ~~
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  Even if you're only interested in using Evocation, not in modifying it, we
encourage you to at least skim through the source. If you've looked at it,
even a little, it won't be so scary next time. It's heavily commented and
meant for anyone with a little programming knowledge to be able to read, even
if you've never done systems programming before.

  If you find something in it confusing, please don't be afraid to ask! It's
likely other people are confused too, and sharing your questions helps improve
the documentation and lets others learn by watching.

  The top-level source file whose job is to compile Evocation itself is
evoke.e. It's really short, and worth a quick glance right now. It lists all
the other source files and the order they get loaded in and how they're
processed.

  The files that do the work to make Evocation run at all are execution.e and
core.e. It's worth reading through both of them slowly. After you've read
core.e, you'll know a lot of basic words that can be used as commands within
Evocation.

  A lot of Evocation is written in Evocation's version of assembly language.
The file amd64.e is the one that implements all the assembly instructions.
Writing a real program in assembly also requires resolving labels, which are
a special syntax that gives names to addresses. The behavior of labels is all
implemented in labels.e.

  On the assembly language front, there's also linux.e which contains assembly
words for doing things specific to the Linux operating system, such as reading
input, and there's elf.e which contains words for outputting the special file
headers that let the operating system understand that a file is an executable
program.

  In terms of the Forth-y bits, input.e and output.e are concerned with
getting text into and out of the language. The infrastructure to define words
is in dynamic.e, and the syntax for it is in interpret.e. The high-level flow
control words are in flow-control.e. Some of the features of execution.e had
to be separated out into their own file, because of details about how the
compiler works; that stuff is in execution-suport.e.

  So, there's all those relatively normal compiler internals in those various
files, which are all fairly self-contained... and then there's the
transformation facility. This is Evocation's most unique architectural
decision, and it's in transform.e. It's well documented, but it's also
extremely conceptually dense. Feel free to give it a skim, that's the only way
to build familiarity with these things, but you should probably have a solid
understanding of the rest of the internals before you place any high
expectations on yourself around understanding the transformation facility.

  It's okay, you can benefit from it before you understand it: Transformation
provides the core tricks that make it possible to compile Forth code into
standalone executables. The call to label-transform in evoke.e, and the call
to log-load-transform in execution.e, are the two spots where compilation is
handed off to the transformation facility, and you can pretty much just take
it for granted that it works, until you feel ready.

  If you want examples of programs that are smaller than Evocation itself,
quine.e is a tiny program written in proper Evocation that outputs its own
source code; hello.e is a hello-world written in Evocation-assembly, and hex.e
is another small Evocation-assembly program that might make a good example of
how to do slightly more complex things that way. All three of these are
self-contained, consisting of just that one file plus calls to Evocation's
built-in library.


 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~ Modifying Evocation's Internals ~~
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  There may come a point in your explorations when you wish to make changes to
the compiler. When you have a change you want to try it out, you can use your
existing copy of Evocation to compile a new one, like this:

  $ cat labels.e elf.e transform.e execution.e evoke.e | ./evoke > evoke2
  $ chmod 755 evoke2

  You can then run ./evoke2 and try out the new features you added. Fun,
right? :)

  If you're planning to submit your changes for inclusion, please also verify
that your compiled output is stable: Run the compilation command again but
change the last part from "./evoke > evoke2" to "./evoke2 > evoke3". The two
versions evoke2 and evoke3 should be bytewise identical; if they are not,
please fix that. This is an important property which would be very difficult
to get back if we ever lose it, it's easier to maintain it in-the-moment.

  For learning about operating system internals and discovering more reference
material, Irenes recommend the osdev wiki[4]. In particular, Evocation's
executable format is ELF[5] in its the 64-bit version[6]; its interface with
the kernel is the System V ABI[7] in its AMD64 version[8]; and the Intel
processor reference manual[9] was consulted heavily for understanding the
instruction set architecture.

  If you want to learn more about Forth implementation in particular, check
out Jonesforth[10]. If you want to see a different creative direction a modern
Forth dialect has gone in, look at Factor[11]. If you're interested in
concatenative languages other than Forth, check out the uxn VM[12] and its
language uxntal[13].

  Irenes sadly do not have recommendations for learning about compiler
concepts.


 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~~ Hexing Evocation for Distribution ~~
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

  The long-term strategy for Evocation's binary bootstrapping is not yet
ready, but it's described here anyway because this is where the explanation
should eventually go, and it's easier to write about the pieces as they're
created.

  The binary bootstrapping strategy rests on something called the
"hex transform", the most complex of the transformations provided as part of
Evocation's transformation facility in transform.e. The hex transform has the
task of transforming an entire compilation process, which would otherwise
produce an executable binary, and instead output a commented hex dump of that
binary which describes its internals and their purpose, byte by byte, in
sufficient detail to allow a human reader to audit their correctness. It will
do this by passing through comments and call-stack information from the
compilation process to the resulting output.

  In order to turn this commented hex dump into a binary, there is a tiny
program called "hex" which handles comments in Evocations ~ syntax, and
converts ASCII hexadecimal to raw binary. This program is in hex.e and is
written in Evocation-assembly. When compiled it is only 480 bytes, which is
small enough to fully audit in its raw, binary form. This is slightly larger
than necessary; many of those bytes are used for error message strings, on the
principle that it's very important that it be easy to distinguish a successful
invocation of "hex" from a failed one.

  The compiled "hex" has proven quite stable, and the hex transform does work
on it. So, a copy of the compiled "hex" is checked into source control so that
it can serve as a root of trust for all Evocation builds. For ease of
auditing, a commented hex dump version of this binary, produced via the hex
transform, is also checked in, as "hex.hex" (We heard you liked
metacircularity, so we put some metacircularity in your metacircularity so you
can be metacircular while you're metacircular.)

  If you need to compile "hex", you can do so as follows:

  $ cat labels.e elf.e hex.e | ./evoke > hex
  $ chmod 755 hex

  To produce the hex-dump version of it, do:

  $ cat labels.e elf.e transform.e inscribe-hex.e | ./evoke > hex.hex

  The program "hex" is written in Evocation-assembly, but the hex transform
also works on Forth programs, including Evocation itself. To produce the
hex-dump version of Evocation, do:

  $ cat labels.e elf.e transform.e inscribe-evoke.e | ./evoke > evoke.hex

  It should run to completion, producing output in evoke.hex. Passing
evoke.hex through ./hex will give a binary that's byte-for-byte identical to
evoke.

  This does take several minutes to run (as many as ten, as of this writing),
but the cause is known: The metadata output buffer is stored in a way that
requires linear traversal for several operations. This dominates the runtime;
when it is fixed, the largest remaining cost will be the use of linked lists
rather than hash tables for the various dictionaries. These are both tasks to
do in the  near future, now that bootstrapping is complete.

  Nearly all of this runtime is attributable to the log-load transform; if
you're working on something that doesn't involve the log-load transform, you
may find it useful to temporarily comment out the call to log-load-transform
in execution.e, replacing it with two invocations of drop. This will not
produce a working compiler, but it will finish faster.

  Now get debugging! :)


 ~~~~~~~~~~~~~~
~~ References ~~
 ~~~~~~~~~~~~~~

  When you read something you enjoy, you should always check its list of
references to discover other stuff you might like. In graduate school they'll
teach you that reading citations is a great way to find out about things, but
here, you just got the lesson for free!

[1] Ken Thompson, "Reflections on Trusting Trust"
    https://doi.org/10.1145/358198.358210
[2] "The Fujoshi Guide to Web Development"
    https://www.fujoweb.dev/
[3] Julia Evans, "wizard zines"
    https://wizardzines.com/
[4] osdev wiki
    https://wiki.osdev.org/
[5] "Tool Interface Standard (TIS) Executable and Linking Format (ELF)
     Specification", version 1.2.
    https://refspecs.linuxfoundation.org/elf/elf.pdf
[6] "ELF-64 Object File Format", version 1.5 draft 2.
    https://uclibc.org/docs/elf-64-gen.pdf
[7] "System V Application Binary Interface"
    https://www.sco.com/developers/devspecs/gabi41.pdf
[8] "System V Application Binary Interface AMD64 Architecutre Processor
     Supplement", version 1.0.
    https://gitlab.com/x86-psABIs/x86-64-ABI/
[9] "Intel 64 and IA-32 Architectures Software Developer's Manual"
    https://www.intel.com/content/www/us/en/developer/articles/technical/intel-sdm.html
[10] https://github.com/nornagon/jonesforth/
[11] https://factorcode.org/
[12] https://100r.co/site/uxn.html
[13] https://wiki.xxiivv.com/site/uxntal.html